改进的稀疏表示遥感图像超分辨重建

为了进一步提高遥感图像超分辨效果,提高超分辨重建速度。针对以往稀疏超分辨算法中更容易丢失边缘信息和引入噪声的问题,本文改进了特征提取算子,以对称近邻滤波(SNN)代替高斯滤波,重点解决特征空间中的字典学习问题。首先,根据遥感图像退化模型生成训练样本图像,并分别对高、低分辨率遥感图像进行7×7分块,生成字典训练样本。然后,建立连接高、低分辨率图像空间的双参数联合稀疏字典,将字典学习过程中的稀疏系数分解为系数权值和字典原子的乘积,依据字典原子指标训练和更新字典,得到高低分辨率联合字典映射矩阵。最后,进行遥感图像超分辨稀疏重构。实验结果表明:与当前最先进的稀疏表示超分辨算法相比,本文算法得到的超分辨重建遥感图像的主观效果更好,恢复出更多的地物细节信息;客观评价参数峰值信噪比(PSNR)提高约1.7dB,结构相似性(SSIM)提高约0.016。改进的稀疏表示超分辨算法可以有效地提高遥感图像超分辨效果,同时降低重建时间。

Remote Sensing Image Super-resolution Based on Improved Sparse Representation

Aim to the problem of losing more details and introducing noise in the previous sparse representation image super-resolution, an improved feature extraction algorithm is proposed to improve the image super-resolution reconstruction (SRR) effect. At the same time, Gaussian filter is replaced by symmetric nearest neighbor filter to speed up image super-resolution and the problem of dictionary learning in feature space is solved. Firstly, a training sample images are generated according to remote sensing image degradation model, and high-low resolution images are respectively divided into image patches size of 7x7. Then, high-low resolution joint dictionary mapping matrix is obtained after the training and updating of dictionary. Finally, image super-resolution reconstruction is performed in sparse representation. Experimental results show that the method proposed in this paper can reconstruct the higher quality super-resolution image with the smaller reconstruction time. At the same time, compared with the most advanced sparse representation super-resolution algorithm, the SRR result image in this paper contains more texture details of ground objects subjectively. PSNR is increased about 1.7dB,and SSIM is increased about 0.016 objectively.